There’s nothing magic about age 27

There are some fantasy baseball analysts who consider age 27 to be a magic number for players. Eric Mack at Sports Illustrated gives his reasoning for the theory:

History shows 27 is the age many players outperform their draft position because a man’s physical peak comes around then and years of preparation allow them to blossom statistically.

It may sound good in theory, but I’m not convinced that history actually shows this to be true. For me, it seems a little simplistic, so I decided to run some numbers and see whether or not it’s true, and if so, to what extent.

The study

My study looks at all players since 1953 (the beginning of the modern baseball era) with at least 400 plate appearances or 130 innings pitched in one season. I then compare each player’s numbers in that season to the year after, grouping by age.

I place no restrictions on the number of plate appearances in the following season because this would introduce a lot of survivor bias. That is, if a player fails to break out and posts poor numbers, it’s likely that his playing time will be reduced.

If we set a plate appearance cutoff, these players would be ignored since they wouldn’t reach the cutoff, which we don’t want. Sure, we’ll get some injuries and some part-time player weirdness mixed in, but I don’t think that’s as important as capturing these legitimate non-breakouts.

After I had my player pool, I decided to define a breakout as a player outperforming his previous year’s production by at least 20 percent. I decided to use four categories: wOBA to measure raw offensive production plus the three main fantasy skill categories for hitters (home runs, batting average and steals). All stats are rate stats, with the rates being HR/(AB-K), H/AB, and SB/AB. For pitchers, I used ERA, WHIP, and strikeouts per nine innings (K/9).

The results

Here is what we get for all ages that had at least 100 player seasons. The numbers list tell us the percentage of players that exceeded their previous year’s production in that category by at least 20%.

+-----+--------+------+-----+-----+-----+
| Age | Sample | wOBA | HR  | BA  | SB  |
+-----+--------+------+-----+-----+-----+
| 22  |    114 |   9% | 39% |  3% | 27% |
| 23  |    256 |   7% | 36% |  7% | 36% |
| 24  |    461 |   5% | 31% |  6% | 35% |
| 25  |    653 |   6% | 36% |  6% | 31% |
| 26  |    854 |   4% | 33% |  5% | 32% |
| 27  |    932 |   2% | 32% |  3% | 28% |
| 28  |    949 |   3% | 29% |  4% | 30% |
| 29  |    902 |   4% | 31% |  5% | 27% |
| 30  |    845 |   3% | 29% |  4% | 25% |
| 31  |    761 |   3% | 29% |  3% | 26% |
| 32  |    660 |   3% | 28% |  4% | 26% |
| 33  |    534 |   2% | 27% |  2% | 23% |
| 34  |    429 |   2% | 29% |  2% | 29% |
| 35  |    331 |   2% | 23% |  3% | 27% |
| 36  |    249 |   2% | 24% |  2% | 24% |
| 37  |    173 |   1% | 21% |  1% | 17% |
| 38  |    110 |   2% | 22% |  1% | 28% |
+-----+--------+------+-----+-----+-----+

Essentially, what we see is that the “age 27″ theory holds little water. Most of the breakouts happen at younger ages (though part of that may be smaller sample sizes), and 27 is essentially like any other mid-20s age in terms of breakout potential. Once you get into the 30s, the likelihood of a breakout starts going down.

If I change the criteria to either 10% above the previous year or 30% above the previous year, the results are very similar, with age 27 not meaning very much.

But what about pitchers?

+-----+--------+-----+------+-----+
| Age | Sample | ERA | WHIP | K/9 |
+-----+--------+-----+------+-----+
| 22  |    102 | 37% |  18% | 14% |
| 23  |    208 | 28% |  14% | 21% |
| 24  |    352 | 32% |  15% | 17% |
| 25  |    457 | 34% |  12% | 17% |
| 26  |    552 | 31% |  14% | 13% |
| 27  |    578 | 32% |  13% | 15% |
| 28  |    550 | 31% |  14% | 15% |
| 29  |    495 | 36% |  15% | 12% |
| 30  |    458 | 34% |  13% | 15% |
| 31  |    369 | 36% |  14% | 10% |
| 32  |    313 | 33% |  17% | 13% |
| 33  |    263 | 36% |  15% | 14% |
| 34  |    209 | 36% |  15% | 19% |
| 35  |    163 | 36% |  15% | 12% |
| 36  |    127 | 32% |  16% | 13% |
| 37  |    100 | 39% |  19% | 12% |
+-----+--------+-----+------+-----+

Nope, same story. Pitchers don’t tend to break out more often at age 27, either.

Initial, ill-fated idea for the study

Originally, I intended to run this study using projections (namely Jeff Sackman’s infinitely useful historical Marcels), comparing a player’s actual performance to his preseason projection. When I ran the tests, however, I got some strange results.

I found that hitters in their 30s were the “most likely to break out,” which seems incredibly counter-intuitive. My best guess as to why is that there’s likely some survivor bias going on that Marcels —being a bare-bones approach to forecasting —isn’t accounting for.

You see, if a hitter lasts into his mid-30s as a major league player, he’s likely of a different breed than a hitter who fizzles out before then. If a player in his 30s plays poorly, he’s less likely to receive another chance than a player who performs poorly in his 20s and still has “potential.”

As an example of this, since 2000, hitters over 30 have posted a 3.9% HR/(AB-K), while hitters under 30 have posted a 2.8% rate (that’s a difference of about five homers over a full season). Marcels doesn’t recognize this, though, and regresses all player stats to the major league average.

But these over-30 hitters aren’t major league average; they’re better than that. Because they’re actually better than what their projections say, they outperform the projections, giving the illusion of a “breakout.”

So, while using projections for this kind of test would have been ideal, Marcels doesn’t seem suited to the task, and no other projection system goes back far enough to give us a large enough sample. If you were curious, though, when compared to other “20s” ages, the age 27 didn’t turn out to be anything special using this Marcels approach.

Concluding thoughts

This is far from a definitive study, but it is one piece of data to consider and certainly doesn’t help the “27 is a magic number” theory. If you have any questions or suggestions, as always, feel free to let me know.

Print Friendly
 Share on Facebook0Tweet about this on Twitter1Share on Google+0Share on Reddit0Email this to someone
« Previous: The reawakening of Vernon Wells
Next: Should I make this trade? »

Comments

  1. Derek Carty said...

    I forgot to add a little disclaimer that I meant to put in there, that being that this isn’t to say that hitters don’t improve until a certain point and then begin to decline once they reach that point.  They do.  And in general, the age 26-29 area is when it happens, depending on what stat you’re looking at.  I’m not disputing aging curves, which are very real and need to be a very real consideration when evaluating players.  But these aging curves are built into projection models, which we have at our disposal.  What I was, in a way, looking for today was evidence that a hitter’s 90th percentile projection is going to be higher for a 27-year-old that it is for any other age because players have a higher chance of breaking out at this age.

    You’re certainly right that my sample peaked at age 26/27, and that’s because raw playing time peaks at age 26/27.  Since 1953, age 26 players have posted a about 850,000 PA and age 27 players have posted 849,000 PA (and this can change a bit depending on how you’re calculating age).  Age 25 is 775,000 and Age 28 is 740,000.

    If I run a quick check on playing time *breakouts*, though, age 27 is nothing special.  In fact, playing time breakouts are relatively linear.  Age 20 sees a 58% playing time breakout and it declines from there.  Age 27 is 26% (down from Age 27: 31% and above Age 28: 24%).  (Note, though, that I just ran this quickly and am not including players who didn’t play in the majors the year before, but I think we’d still find the same thing, maybe skewed even more on the younger side, since most minor leaguers get called up and would get that 100% playing time breakout rate when they’re in their early 20s.)

    So if you take issue with me using rates to define breakouts in the article, we see here that even using just playing time, we don’t see any breakouts at age 27.

    Injuries would take a little more time to check out, and you might be right here, but we also need to make sure that we’re asking the right questions.

    The question is not “do 27-year-olds get injured less frequently then 26 year olds?”, just as the initial question was not “do 27-year-olds perform better than 26-year-olds.”  Yes, overall they probably get injured less, but the real question is “do they get injured less relative to their previous injuries at a rate lower than 26-year-olds do (or any other age).”  And I don’t think it’s easy to say that they do, at least without evidence.

    That’s because what we’re looking for here is information above and beyond what we already know.  We already know a player’s past performance and injury history.  We want to know if his age being 27 makes him any more likely to outperform previous levels or get injured less than if he was a different age.

  2. Ameer said...

    Not to be a party-pooper, but didn’t the original Bill James study on this describe a 5-year average peak between ages 25 and 29?  I know that it’s commonly stated that hitters “break out” at 27, but the original essay never stated this.  An average peak and a “breakout” are very, very different things.

    I like the article, but you might be refuting something that was never actually claimed to be true (although the age 27 breakout has become one of those weird baseball truisms over time).

  3. Derek Carty said...

    Yeah, Ameer, I think you’re mostly spot on.  I’m not talking about aging curves as I imagine James was.  Those are real and generally the peak comes between 25 and 29, depending on the stat, but as you said, and average peak and a breakout are different things.

    If you look at the link at the top to SI, though, that’s the main place I’ve seen this “age 27 breakout” theory supported.  It’s also been at CBS (by the same author) in past years.  I’m sure some people don’t buy into it, but others do, so I thought I’d take a bit more methodical look at it.  And as you said, it has sort of become of those “weird baseball truisms” that has sort of become accepted by some, so I wanted to look at it a but more objectively.

  4. Ameer said...

    Ahhh, that’s what I get for not clicking on the link.  I didn’t even see that the first time I read it.

  5. Dan Creps said...

    Your sample peeked at 27-28, which seems to support the theory, I still believe you have better stats, less injuries around the 27-28 time frame, and in fantasy, injuries can kill you.

  6. Derek Carty said...

    You’re right, Ed.  I just did a quick Google search and came up with 1953 (because of the Braves moving), and I thought I’d seen that number used before, but that doesn’t appear to be the right definition of modern baseball era upon further searching.  I’m not a baseball historian by any means, so my apologies.  Maybe 1961 or 1977 would have been better to use, though in some ways it’s arbitrary.

    If I re-run the numbers using 1961, I get the same end result, though.

  7. Peter Kreutzer said...

    Hi Derek,

    I did a very similar study some years back, using my historical roto prices for comparison. I defined the “modern era” as ML careers that started after WWII and included all players who were older than 35 in 1999.

    But I didn’t look at player’s jumps from the previous season, focusing rather at the age they earned the most roto dollars (this aggregates the four roto cats, and adjusts production to league context, so it smooths out shifts in context) in their career.

    Just as you found, sample size peaked at 26/27, but the rate of players having peak years at each age was higher for the younger players and declined as players aged (though the average production per player at each age stayed pretty constant right up until retirement age).

    But because there were more 26 and 27 year olds playing it was players at that age who produced the most fantasy value by far. Your numbers suggest as much, too.

    Apropos emack’s statement, which launched your survey, I think the numbers suggest that it is the players who debut at 26 or 27 who are benefiting from the synergy of physical peak meeting mental peak. These are the guys who debut late and often erupt out of nowhere, and then don’t usually stick around long. That’s the profile age 27 watchers should keep an eye on.

  8. GTWMA said...

    This age 27 “myth” is one commonly repeated on fbb sites and boards, and while Derek’s look may not be the definitive study, it shows what all the other studies show—age 27 is not a magic number.  While it’s the most common peak, plenty of players peak at other ages between 24-30, and there’s nothing dramatic about the peak at age 27.

  9. Ed Buskirk Jr. said...

    1953 is the beginning of the modern baseball era? I’ve never heard that before, and I’d like to know your reasoning. MLB says the modern era started in 1901, some historians say 1961, and some yahoo at Yahoo! Sports seems to think it started in 1969, and I can think of actual reasons for each of those years. But 1953? The only major rule change was reinstating the sacrifice fly rule. Please don’t tell me that the Braves moving marked the beginning of the modern era.

  10. Jimbo said...

    I wonder if it would be possible to systematically identify when a player’s “career year” occurs. Probably is, just depends how one defines a career year eh?

    What I take the ‘age 27’ myth to mean is that the chance of a player’s BEST season coming at that age is highest.

    For fantasy, I think downside is as important as upside. Without seeing any names, if you gave me the option of a team full of 23-25 year olds, 26-28 year olds, and 29-32 year olds…I’ll take the middle option. Odds are best for ‘peak production’ and probably has the least amount of risk as well.

    This year I focused heavily on prime years. Only Stanton and Dunn are outside of the 26-28 range (2 of 23 players).

  11. Brian Cartwright said...

    I thought you might have used 1953 because it was the start of the Retrosheet era (but that gets pushed back a little more each season, now I believe it’s 1950).

    For a little bit of shameless self-promotion, Oliver does regress based on age, pos and level. Pedro Alvarez is regressed to the recent historical mean of 23 year old MLB 3rd basemen, while Lyle Overbay is regressed to 33 year old MLB 1st basemen.

  12. Dave Studeman said...

    My memory is that Ron Shandler really popularized the notion of players “popping” during the age-27 years.  He used to give advice that basically said that you should bid extra for players at age 27.

    My memory is not what it used to be, but I have always associated the “extra oomph” theory to Ron.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Current ye@r *